February 2012 - Libguestfs - libguestfs List Archives

Re: [Libguestfs] want to access the page file of VM from host on Xen platform using libguestfs
by Richard W.M. Jones 10 Feb '12

10 Feb '12

On Thu, Feb 09, 2012 at 03:17:22PM -0800, Irfan Ahmed wrote: > I am using Xen and want to access the page file contents of windows > XP SP2 virtual machine from dom0. Is it possible to read the page > file of live Windows XP virtual machine using libguestfs. I have a C > program and want to extend its functionality using > libguestfs. Please help me how this can be done. Yes; this can be done by reading the file through libguestfs, which is what 'virt-cat' (a C program) does. I would suggest you start with these two examples: http://libguestfs.org/guestfs-examples.3.html https://github.com/libguestfs/libguestfs/blob/6ca8a2db6d6f1028a35ece616af44… If you need to locate the page file (ie. it's not in a default location) then you will have to read that from the registry, which libguestfs can also do. If the guest is live you won't simply be able to keep a handle open and re-read the file. You might need to map the file to disk blocks and read them directly. (This is a bit more experimental) Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones virt-top is 'top' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://et.redhat.com/~rjones/virt-top

2 2

[PATCH 0/3] Fix guestfish edit command.
by Richard W.M. Jones 10 Feb '12

10 Feb '12

This is a further, more comprehensive fix for https://bugzilla.redhat.com/show_bug.cgi?id=788641 The guestfish 'edit' command (aka 'emacs', 'vi') suffered from the same problems as virt-edit and more. It could have failed and left a partially overwritten file, and it didn't preserve permissions etc from the original file. These three patches fix all this. The first is a code cleanup. The second causes 'edit' to write to a temporary file which is atomically moved over the original. The third copies permissions etc from the original to the new file. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones virt-df lists disk usage of guests without needing to install any software inside the virtual machine. Supports Linux and Windows. http://et.redhat.com/~rjones/virt-df/

1 3

[PATCH-v4 0/2] Add a new api wipefs
by Wanlong Gao 10 Feb '12

10 Feb '12

Hi Rich: This is a latest patch. ;-P Comments? Thanks -Wanlong Gao Wanlong Gao (2): NEW API: add a new api wipefs virt-format: erase the filesystem signatures before erase partitions TODO | 2 -- daemon/zero.c | 17 +++++++++++++++++ format/format.c | 19 ++++--------------- generator/generator_actions.ml | 14 ++++++++++++++ src/MAX_PROC_NR | 2 +- 5 files changed, 36 insertions(+), 18 deletions(-) -- 1.7.9

2 4

want to access the page file of VM from host on Xen platform using libguestfs
by Irfan Ahmed 09 Feb '12

09 Feb '12

I am using Xen and want to access the page file contents of windows XP SP2 virtual machine from dom0. Is it possible to read the page file of live Windows XP virtual machine using libguestfs. I have a C program and want to extend its functionality using libguestfs. Any idea how this can be done using libguestfs. --Irfan

1 0

[PATCH v3 0/2] Add a new api wipefs
by Wanlong Gao 09 Feb '12

09 Feb '12

Hi Rich: This is a alternative patch, if you like the api name called wipefs ;-P Thanks -Wanlong Gao Wanlong Gao (2): NEW API: add a new api wipefs virt-format: erase the filesystem signatures before erase partitions TODO | 2 -- daemon/zero.c | 17 +++++++++++++++++ format/format.c | 19 ++++--------------- generator/generator_actions.ml | 14 ++++++++++++++ src/MAX_PROC_NR | 2 +- 5 files changed, 36 insertions(+), 18 deletions(-) -- 1.7.9

3 6

Fix virt-edit so it preserves permissions (RHBZ#788641)
by Richard W.M. Jones 08 Feb '12

08 Feb '12

The first patch preserves file mode, UID, GID and SELinux context across edited files. The second patch adds a useful new command in guestfish ('llz') which shows SELinux context (like 'ls -laZ') that was useful when debugging this. Rich.

1 2

[PATCH 1/3] NEW API: add a new api zero_fs
by Wanlong Gao 07 Feb '12

07 Feb '12

Add the new api zero_fs to erase the filesystem signatures on a device but now erase any data. Signed-off-by: Wanlong Gao <gaowanlong(a)cn.fujitsu.com> --- daemon/zero.c | 17 +++++++++++++++++ generator/generator_actions.ml | 14 ++++++++++++++ src/MAX_PROC_NR | 2 +- 3 files changed, 32 insertions(+), 1 deletions(-) diff --git a/daemon/zero.c b/daemon/zero.c index 398f844..348869f 100644 --- a/daemon/zero.c +++ b/daemon/zero.c @@ -73,6 +73,23 @@ do_zero (const char *device) } int +do_zero_fs (const char *device) +{ + int r; + char *err = NULL; + + const char *wipefs[] = {"wipefs", "-a", device, NULL}; + r = commandv (NULL, &err, wipefs); + if (r == -1) { + reply_with_error ("%s", err); + free (err); + return -1; + } + free (err); + return 0; +} + +int do_zero_device (const char *device) { int64_t ssize = do_blockdev_getsize64 (device); diff --git a/generator/generator_actions.ml b/generator/generator_actions.ml index 3a7be79..d27ef43 100644 --- a/generator/generator_actions.ml +++ b/generator/generator_actions.ml @@ -6651,6 +6651,20 @@ This option may not be specified at the same time as the C<correct> option. =back"); + ("zero_fs", (RErr, [Device "device"], []), 305, [Progress], + [InitBasicFSonLVM, Always, TestRun ( + [["zero_fs"; "/dev/VG/LV"]])], + "wipe a filesystem signature from a device", + "\ +This command can erase filesystem or raid signatures (magic strings) from +the specified C<device> to make the filesystem invisible for libblkid. +This does not erase the filesystem itself nor any other data from the C<device>. +Compare with C<guestfs_zero> which just zeroes the first few blocks of a +device, contained the partitions on the device, but this C<guestfs_zero_fs> +can't erase the partitions of a device. + +=back"); + ] let all_functions = non_daemon_functions @ daemon_functions diff --git a/src/MAX_PROC_NR b/src/MAX_PROC_NR index 873b744..67d04b9 100644 --- a/src/MAX_PROC_NR +++ b/src/MAX_PROC_NR @@ -1 +1 @@ -304 +305 -- 1.7.9

2 6

[PATCH 1/2] Revert "daemon: Run udev_settle after pwrite-device finishes."
by Richard W.M. Jones 06 Feb '12

06 Feb '12

From: "Richard W.M. Jones" <rjones(a)redhat.com> This reverts commit a9c8123c72db47bcab8dd738e8d5256a9ae87f11. --- daemon/file.c | 18 +++--------------- daemon/parted.c | 3 +-- 2 files changed, 4 insertions(+), 17 deletions(-) diff --git a/daemon/file.c b/daemon/file.c index 057e15d..91746e0 100644 --- a/daemon/file.c +++ b/daemon/file.c @@ -525,7 +525,7 @@ do_pread_device (const char *device, int count, int64_t offset, size_t *size_r) static int pwrite_fd (int fd, const char *content, size_t size, int64_t offset, - const char *display_path, int settle) + const char *display_path) { ssize_t r; @@ -541,18 +541,6 @@ pwrite_fd (int fd, const char *content, size_t size, int64_t offset, return -1; } - /* When you call close on any block device, udev kicks off a rule - * which runs blkid to reexamine the device. We need to wait for - * this rule to finish running since it holds the device open and - * can cause other operations to fail, notably BLKRRPART. 'settle' - * flag is only set on block devices. - * - * XXX We should be smarter about when we do this or should get rid - * of the udev rules since we don't use blkid in cached mode. - */ - if (settle) - udev_settle (); - return r; } @@ -575,7 +563,7 @@ do_pwrite (const char *path, const char *content, size_t size, int64_t offset) return -1; } - return pwrite_fd (fd, content, size, offset, path, 0); + return pwrite_fd (fd, content, size, offset, path); } int @@ -593,7 +581,7 @@ do_pwrite_device (const char *device, const char *content, size_t size, return -1; } - return pwrite_fd (fd, content, size, offset, device, 1); + return pwrite_fd (fd, content, size, offset, device); } /* This runs the 'file' command. */ diff --git a/daemon/parted.c b/daemon/parted.c index 64a7d1d..16f0843 100644 --- a/daemon/parted.c +++ b/daemon/parted.c @@ -34,8 +34,7 @@ * COMMAND_FLAG_FOLD_STDOUT_ON_STDERR flag. * * parted occasionally fails to do ioctl(BLKRRPART) on the device, - * probably because udev monitors all 'close' on block devices - * and runs 'blkid' which opens and examines the device. We attempt + * apparently because of some internal race in the code. We attempt * to detect and recover from this error if we can. */ static int -- 1.7.6

1 2

[PATCH 0/2] Work-around blkid running from udev after device close.
by Richard W.M. Jones 06 Feb '12

06 Feb '12

This is a possibly better work-around for bug 769304, the bug where blkid runs from udev whenever a device closes, causing the subsequent operation to fail. For background to this change, see: https://rwmj.wordpress.com/2012/01/19/udev-unexpectedness/#content So far we have noticed this problem in two places (although I'm fairly certain it affects many more): - in virt-resize, after copy_device_to_device and before part_set_bootable (which runs parted) - in virt-format, after wiping the device (which writes zeroes then closes it) and before blockdev_rereadpt The fix is to run udev_settle *before* all blockdev, parted and sfdisk commands. Note we do it before because at this point a blkid command may be running which we have to wait for. None of these commands should be performance sensitive (unlike pwrite_device, where I have removed the udev_settle call). In the parted and sfdisk cases we also have to run this command afterwards, since parted itself will have closed the block device. There may be alternatives to this patch: - Add a udev rule (temporarily?) which causes udev to ignore devices that libguestfs is using. However I feel this could be dangerous since it may affect more than just the problematic blkid operation. - Edit the udev rules so that the blkid rule doesn't run. This requires complex and distro-specific edits to built-in udev rules. - Get something into udev upstream which lessens the effect of the blkid rule (eg. delaying it or disabling it in some circumstances) Rich.

1 2

[PATCH] build: Choose a virtual directory for the daemon which is not a symlink.
by Richard W.M. Jones 06 Feb '12

06 Feb '12

From: "Richard W.M. Jones" <rjones(a)redhat.com> Check /sbin, /usr/sbin, /bin and /usr/bin to ensure that neither the directory nor the parent (eg. /usr) is a symlink, and use this directory for the virtual directory used to store guestfsd in the appliance. --- appliance/Makefile.am | 9 +++++---- configure.ac | 22 ++++++++++++++++++++++ 2 files changed, 27 insertions(+), 4 deletions(-) diff --git a/appliance/Makefile.am b/appliance/Makefile.am index 99896be..fcf6422 100644 --- a/appliance/Makefile.am +++ b/appliance/Makefile.am @@ -67,10 +67,11 @@ stamp-supermin: make.sh packagelist excludelist supermin.d/daemon.img: ../daemon/guestfsd mkdir -p supermin.d rm -f $@ $@-t - mkdir sbin - cd sbin && ln ../../daemon/guestfsd - echo -e "sbin\nsbin/guestfsd" | cpio --quiet -o -H newc > $@-t - rm -r sbin + rm -rf tmp + mkdir -p tmp/$(DAEMON_SUPERMIN_DIR) + cd tmp/$(DAEMON_SUPERMIN_DIR) && ln $(abs_top_builddir)/daemon/guestfsd + ( cd tmp && find | cpio --quiet -o -H newc ) > $@-t + rm -rf tmp mv $@-t $@ supermin.d/init.img: init diff --git a/configure.ac b/configure.ac index 5caca32..019f7e4 100644 --- a/configure.ac +++ b/configure.ac @@ -278,6 +278,28 @@ if test "x$enable_daemon" = "xyes"; then AC_DEFINE([VALGRIND_DAEMON],[1],[Define to 1 to run the daemon under valgrind]) AC_DEFINE_UNQUOTED([VALGRIND_LOG_PATH],["$(pwd)"],[Path to save valgrind log files]) fi + + dnl Which directory should we put the daemon in? NOTE: This + dnl is the "virtual" directory inside the appliance, not the + dnl install directory for libguestfs live. Since Fedora 17 + dnl /sbin is a symlink to /usr/sbin. We have to put the + dnl daemon into a real (non-symlink) directory. + dirs="/sbin /usr/sbin /bin /usr/bin" + AC_MSG_CHECKING([which of $dirs is a real directory]) + for dir in $dirs; do + parent=`dirname $dir` + if test \! -L $parent && test -d $parent \ + && test \! -L $dir && test -d $dir + then + DAEMON_SUPERMIN_DIR=$dir + break + fi + done + if test "x$DAEMON_SUPERMIN_DIR" = "x"; then + AC_MSG_ERROR([non-symlink binary directory not found]) + fi + AC_MSG_RESULT([$DAEMON_SUPERMIN_DIR]) + AC_SUBST([DAEMON_SUPERMIN_DIR]) fi AM_CONDITIONAL([INSTALL_DAEMON],[test "x$enable_install_daemon" = "xyes"]) AM_CONDITIONAL([VALGRIND_DAEMON],[test "x$enable_valgrind_daemon" = "xyes"]) -- 1.7.6

1 0